English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Debiasing methods that seek to mitigate the tendency of Language Models (LMs)to occasionally output toxic or inappropriate text have recently gainedtraction. In this paper, we propose a standardized protocol which distinguishesmethods that yield not only desirable results, but are also consistent withtheir mechanisms and specifications. For example, we ask, given a debiasingmethod that is developed to reduce toxicity in LMs, if the definition oftoxicity used by the debiasing method is reversed, would the debiasing resultsalso be reversed? We used such considerations to devise three criteria for ournew protocol: Specification Polarity, Specification Importance, and DomainTransferability. As a case study, we apply our protocol to a popular debiasingmethod, Self-Debiasing, and compare it to one we propose, called InstructiveDebiasing, and demonstrate that consistency is as important an aspect todebiasing viability as is simply a desirable result. We show that our protocolprovides essential insights into the generalizability and interpretability ofdebiasing methods that may otherwise go overlooked.

Debiasing should be Good and Bad: Measuring the Consistency of Debiasing Techniques in Language Models